Your llama-server configurations, saved as named recipes. Switch models and GPU layouts in one click โ no flag juggling.
Stop memorizing flags. Start managing recipes.
Save each llama-server invocation as a named recipe. Switch between Mistral, Llama, and CodeLlama setups with a single click.
No account. No telemetry. No network calls. Every recipe lives in a SQLite database on your machine โ yours, and yours alone.
A flag deny-list blocks arguments that conflict with app-managed settings. Model paths are confined to your configured directory.